Classification of Healthy and Unhealthy Abaca leaf using a Convolutional Neural Network (CNN)

Abstract

Early suppression or identification of Abaca plant diseases is one of the difficulties for the farmers in Abaca fields, relaying only to the manual process of identifying Abaca plant diseased which were lack of time efficiency and feasibility solution that can cause widespread outbreaks of the diseased Abaca plants. But through the help of the system using the Raspberry Pi 4 and the Raspberry Pi HQ camera, the developed prototype can identify the healthy and unhealthy leaves through the deep learning algorithm of the CNN upon the architecture method ResNet50. The system trained over 200 images sample through the gathered data by the researcher with two classification sets of images consisting of 100 healthy leaves and 100 unhealthy leaves samples under the approval and labeled by the PhilFIDA Catanduanes. The researcher manually took the data sets on the Abaca leaves from the Abaca plantation area in Barangay San Miguel Baras Catanduanes. The thorough division of the Abaca leaf training model by the CNN – ResNet50 and the accuracy training and validation rate reached 100%. The precision rate of the two-output data classification reached 100%.